Minimization of useless work in resource failure recovery of workflow schedules

Abstract

Real-life scheduling has to face many difficulties such as dynamics of manufacturing environments with unforeseen events occurring during the execution of a schedule. Namely, in the case of a resource failure, it may be necessary to process a lot of work again, or a feasible schedule recovery may not exist at all. Moreover, the time window within which the ongoing schedule must be updated may be very short, and too timeconsuming computation of the schedule may lead to a failure of the scheduling mechanism and setback in production. Our approach in the area of predictive-reactive scheduling is to allow for substitution of tasks, which cannot be executed, with a set of alternative tasks. This paper makes use of the model of the hierarchical workflows and gives an SMT and a CSP models to recover an ongoing schedule from a resource failure with the objective to minimize the work processed in vain. The experimental analysis identified parameters for which the SMT model clearly outperforms the CSP model and vice versa.